Reducing time code takes to run using profile viewer

This code below takes up about 40 seconds when the size of the matrix is (13,5,2000) All im doing is removing any matrixs which are repeated.

ie if big_data(:,:,x)==big_data(:,:,x+1) then remove one of the matrixs. I know there is a unique function, but I cant seem to get it to work how I want. When I scale it up to matrixs of size (13,5,10000), its just taking too long.

Is there a better way? thanks.

while x<length(big_data(1,1,:))+1

        while y<length(big_data(1,1,:))        
            y=y+1;

            if big_data(:,:,x)==big_data(:,:,y)
                big_data(:,:,y)=[];
                y=y-1;
            end       
        end

        x=x+1;
        y=x;
    end

Answer 1

[Jan]

Shrinking a large array iteratively requires a lot of resouerces, equivalently to growing. Better use a logical vector to mark subarrays and delete them all together after the loops.

Answer 2

[Paul]

Are you saying to do something like, replace

big_data(:,:,y)=[];

with

to_delete(count)=y

then once the loop is finished

big_data(:,:,to_delete)=[]

[Sean de Wolski]

yup.

But be sure to preallocate to_delete.

Or even better use a logical vector:

to_delete = false(1,1,size(big_data,3));

while ...

to_delete(1,1,index) = true

end

etc